I'm working on a deep learning project and I got stucked: I have a jupiter notebook that processes a dataset of images and returns a list of tensors in python, which I pickle with pickle.dump(). Before, I was unpickling the list in pyhon, and then using it to train my model, so I had no issues... But now I need those pickled data loaded also in the browser (to be able to use them with tensorflow js). Do you know any good alternative to pickle? Obviously I could save my data in a different format in python, so that after I'd be able to load them in js... I just don't know which format is the best (JSON?).

To be clearer on the type of data, this is my data before pickling/after unpickling it in python:

tf.Tensor(
[[[[ 31.741358  36.196915  36.969135]
   [ 31.762732  40.762733  39.762733]
   [ 28.562115  38.562115  37.562115]
   ...
   [ 21.795372  27.795372  27.795372]
   [ 29.314323  33.314323  32.314323]
   [ 25.28419   29.28419   28.28419 ]]

  [[ 33.765278  39.765278  39.765278]
   [ 31.849304  40.849304  39.849304]
   [ 29.36111   39.36111   38.36111 ]
   ...

Thank you in advance!